Optimizing Crop Rotation for Small-scale Farmers in Rwanda

Check the STREAMLIT APP

Overview

This project focuses on developing an optimized crop rotation system for small-scale farmers in Rwanda, using a combination of data-driven analysis and machine learning models to make informed crop recommendations and finally to create a crop rotation plan. This approach is intended to improve soil health, increase crop yield, and support sustainable agricultural practices.

Objectives

• Recommend crop rotations tailored to soil and climate conditions.
• Enhance long-term soil quality while supporting higher yields and profitability.
• Use genetic algorithms to design robust, sustainable crop rotation strategies.
• App development for practical use by stakeholders.

Dataset Overview

The dataset soil.impact.csv provides information on various crops along with their environmental and soil requirements. The dataset includes columns such as:

• Name: Crop name (e.g., Strawberry, Potato)
• Fertility: Fertility level required for crop growth
• Temperature: Temperature (°C) associated with optimal crop growth
• Rainfall: Rainfall levels (mm) suitable for crop growth
• pH: Soil pH requirements
• Light Hours/Intensity: Required daily sunlight hours and light intensity
• Rh: Relative Humidity (%)
• Nitrogen, Phosphorus, Potassium: Key soil nutrients
• Yield: Expected crop yield
• Soil Type: Type of soil best suited for each crop (e.g., Loam)
• Season: Recommended growing season (e.g., Summer, Spring)
• Impact: Soil impact status (e.g., "depleting" indicating nutrient depletion) - new column created in the Feature engineering part

Notebook Structure

1. Data Loading and Initial Exploration

   • Libraries are imported, and the dataset is explored using tools like skimpy for summarization and distribution insights.

2. Exploratory Data Analysis (EDA)

   • Visualizations and statistical summaries to understand correlations and distributions.

3. Data Preprocessing

   • Preprocessing steps include label encoding, scaling, and train-test splitting to prepare for model training.
   • Key libraries used: pandas, numpy, and scikit-learn.

4. Modeling for Crop Recommendation

   • Implementing multiple machine learning models to predict crop suitability based on soil characteristics and environmental factors:
     • GradientBoostingClassifier
     • RandomForestClassifier
     • LogisticRegression
     • XGBoost (XGBClassifier)
   • Evaluation includes confusion matrix and feature importance to choose the final model.

5. Genetic Algorithm for Crop Rotation Optimization

   • Implements a genetic algorithm to generate and evolve crop rotation plans.
   • The algorithm optimizes rotations for 2-5 years, balancing crop yields, and soil health. With market data it can be added profitability as a feature to (also) maximize.
   • Evaluates solutions based on constraints like nutrient cycling and environmental sustainability.

Streamlit App

• Crop recommendation- user input: environmental and soil features.
• Crop rotation with Genetic Algorithms for 2-5 years, 1-4 crops to grow and soil type.